Negative interactions determine Clostridioides difficile growth in synthetic human gut communities

Abstract Understanding the principles of colonization resistance of the gut microbiome to the pathogen Clostridioides difficile will enable the design of defined bacterial therapeutics. We investigate the ecological principles of community resistance to C. difficile using a synthetic human gut microbiome. Using a dynamic computational model, we demonstrate that C. difficile receives the largest number and magnitude of incoming negative interactions. Our results show that C. difficile is in a unique class of species that display a strong negative dependence between growth and species richness. We identify molecular mechanisms of inhibition including acidification of the environment and competition over resources. We demonstrate that Clostridium hiranonis strongly inhibits C. difficile partially via resource competition. Increasing the initial density of C. difficile can increase its abundance in the assembled community, but community context determines the maximum achievable C. difficile abundance. Our work suggests that the C. difficile inhibitory potential of defined bacterial therapeutics can be optimized by designing communities featuring a combination of mechanisms including species richness, environment acidification, and resource competition.

Model prediction OD600 Experiment calculated OD600 Figure EV1. Analysis of parameter uncertainty and predictive capability of generalized Lotka-Volterra models.
A Scatterplot of goodness of fit of experimental absolute abundance (calculated OD600) versus simulated species absolute abundance using the Full Model for the communities in the training data set (Pearson r = 0.89, P = 0.0). Error bars represent one SD from the mean of biological replicates. Gray line indicates y = x, or 100% prediction accuracy. Calculated OD600 is the product of 16S relative abundance and community OD600. B Swarmplot highlighting 24 communities chosen as the held-out set (data also shown in Fig 2A). Orange datapoints represent held-out communities from training set.
Black datapoints indicate communities from Fig 2A in training data set. Calculated OD600 is the product of 16S relative abundance and community OD600. C Scatterplot of experimental absolute abundance (calculated OD600) versus predicted species absolute abundance (OD600) using the Preliminary Model for the 24 held-out communities (Pearson r = 0.52, P = 1*10 −14 ). Error bars represent one SD from the mean of biological replicates. Gray line indicates y = x, or 100% prediction accuracy. Calculated OD600 is the product of 16S relative abundance and community OD600.   Saturates (OD 48hr -OD 12hr < 0.025 in HD or LD) Figure EV3. Dependence of C. difficile abundance on propagule pressure over time.
A Lineplots of C. difficile (CD) abundance over time in CommA-CommO communities. Final timepoint is same as shown in Fig 3A. In low-density conditions, C. difficile inoculated at 10% of total community OD600 at 0 h. In high-density conditions, C. difficile inoculated at 65% of total community OD600 at 0 h. Data points indicate biological replicates and lines indicate mean value of biological replicates. Communities where C. difficile abundance saturates by 48 h in either highdensity or low-density conditions are highlighted in green (saturation defined as difference in C. difficile OD600 between 48 and 12 h is less than 0.025). B-D Lineplots of predicted absolute abundance (OD600) of C. difficile at 12, 24, and 96 h as a function of the initial fraction of C. difficile as simulated by the Full Model.
Data information: In A, n = 1-3 biological replicates (See Appendix Table S4 for replicate information of each condition).  Figure EV4. Impact of C. difficile on resident species abundances.
A-C Heatmap of the fold change of species absolute abundance (mean value of biological replicates) in full community with 5-60% initial C. difficile (CD) compared to the 0% initial C. difficile condition. Stars represent statistical significance: *P < 0.05, **P < 0.01, ***P < 0.001 according to an unpaired t-test. A: C. difficile strain MS002, B: C. difficile strain MS010, C: C. difficile strain MS011. D-F Lineplots of species absolute abundance (calculated OD600) at 48 h as a function of initial C. difficile fraction. Datapoints indicate biological replicates and lines indicate the mean. Calculated OD600 is the product of 16S relative abundance and community OD600. Stars indicate a statistically significant difference in the absolute abundance of B. hydrogenotrophica compared to the absolute abundance of B. hydrogenotrophica in 0% initial C. difficile condition: *P < 0.05, **P < 0.01, ***P < 0.001, ns = no significant difference according to an unpaired t-test. D: CommP, E: CommQ, F: CommR.
Data information: In A-C, D, F n = 3 biological replicates. In E, n = 1 or n = 3 biological replicates (see Appendix Table S4 for replicate information of each condition). A Lineplots of monospecies growth as a function of the initial environmental pH of adjusted fresh media. Growth is quantified as area under the curve (AUC) from 0 to 20 h. Datapoints indicate biological replicates and dashed lines indicate linear regression fits. B Barplot of slopes of linear regression fit to data in A. Stars denote statistical significance: *P < 0.05, **P < 0.01, ***P < 0.001 according to an unpaired t-test. C Scatterplot of decay constants from Fig 2F as a function of pH sensitivity slopes in B. Linear regression y = 0.06x+0.05, Pearson r = 0.44, P = 0.11).
Data information: In A, n = 2-3 biological replicates (See Appendix Table S4 for replicate information of each condition).